1. Field of the Invention
The invention relates to optical flow measurement or computation and, more particularly, to an optic flow sensor incorporating an improved photoreceptor array.
2. Description of the Related Art
As set forth in my earlier U.S. Pat. No. 6,020,953, the subject matter of which is incorporated herein by reference, the term xe2x80x9coptical flowxe2x80x9d generally refers to the motion of texture seen by an agent (such as an animal or a robot) as a result of relative motion between the agent and other objects in the environment. It is well known that animals, especially insects, use information from the optical flow for depth perception and to move about in an environment without colliding into obstacles. Robotics and machine vision researchers have borrowed from these ideas in biology to build matching vision systems that successfully use optical flow for depth perception and obstacle avoidance. These successes verify that optical flow can indeed be used for depth perception and obstacle avoidance in real systems. In a xe2x80x9cneuromorphicxe2x80x9d approach, computations are performed with analog or mixed-mode circuitry which exploit the physical dynamics inherent in VLSI circuitry and often mimic biological structures.
In the above-identified patent, there is disclosed a feature tracking linear optic flow sensor which includes photoreceptor array which responds to light from the visual field focussed thereon. In my U.S. Pat. No. 6,194,695 B1, the subject matter of which is also incorporated by reference, a number of different ways are discussed of implementing a photoreceptor array for linear optic flow measurement. By using a combination of electronics and optics, the individual photoreceptors have visual field response functions that are shaped like a fuzzy elongated rectangle. These photoreceptors are also arranged in a linear array so that the direction of the array (the sensor orientation vector or SOV) is perpendicular to the long axis of that rectangle. The patent discusses two general ways in which the elongated photoreceptor receptive field can be created. The first way is to use a phototransistor (photoreceptor) the active area of which is rectangular. In this embodiment, the photoreceptors of the array are located on a focal plane chip. A lens or pinhole focuses an image of the environment or visual field onto the focal plane chip. Thus, the photoreceptors sample a rectangular section of the image, and, therefore, of the visual field. This rectangular response function can be made slightly blurry through the use of optical smoothing by either placing the lens slightly out of focus or by using an iris with a graded transmission function.
The second way to achieve an elongated rectangular photoreceptor receptive function is to use photoreceptors the active areas of which are substantially point-like and to also use an iris with an elongated rectangular shape. With such an arrangement, all the light striking the photoreceptor will be from an elongated rectangular part of the visual field.
Each of these two ways of forming elongated photoreceptor receptive fields has a potential shortcoming. The main shortcoming of the first pinhole-camera version is that a very small amount of light strikes the photoreceptor. This results in a very small current flowing through the photoreceptor. Because of this, the parasitic capacitance between the photoreceptor (phototransistor) and ground takes more time to charge or discharge. This effectively slows down the photoreceptor so that only slowly changing textures can be detected. Faster intensity changes due to faster optic flow are merely filtered out. This same effect is observable in the second version of the sensor in which point photoreceptors are used with rectangular irises. However, the effective cutoff frequency is higher because more light is let in by the rectangular iris and less parasitic capacitance needs to be overcome in the phototransistors, which are physically smaller.
Considering the above-mentioned shortcoming in a more rigorous fashion, reference is made to FIGS. 1 and 2 wherein FIG. 1 shows a minimal photoreceptor circuit for analysis and FIG. 2 shows a linearized version of the photoreceptor circuit for AC analysis. As shown in FIG. 1, the basic photoreceptor circuit 10 consists of a PNP phototransistor 12 and a diode-connected MOSFET 14. The phototransistor 12 connects to ground (or the substrate) an amount of current proportional to the total light striking its surface. The diode-connected MOSFET 14 converts the photoreceptor current into a voltage. Typically MOSFET 14 is wide enough so that it is biased in the subthreshold region. When a diode-connected MOSFET is in the subthreshold region, the current flowing through it is of the form
Idiode=Is exp(kVgs)xe2x80x83xe2x80x83(Eq. 1)
where Is is dependent on the MOSFET geometry (and fabrication process), Vgs is the gate to source voltage, and k is dependent on the fabrication process. Thus, the voltage drop across MOSFET 14 is a logarithmic function of the phototransistor current. Typical k values are on the order of 50-100 mV. Accordingly, as a result of the logarithmic compression, several orders of magnitude of light intensity can be handled by the photoreceptor circuit 10.
As indicated above, FIG. 2 shows a linearized version of the photoreceptor circuit for a specific operating phototransistor current loop. This linearization is performed in exactly the same manner as is used to analyze simple transistor amplifier circuits. The phototransistor is modeled as a current source Iin in parallel with a capacitance Cp. The current source iin represents or refers to the deviation in current from the operating point current Idiode. The capacitance Cp represents or refers to the parasitic capacitance between the phototransistor and the substrate. The diode-connected MOSFET 14 can be reduced to a conductance gm, since the gate is connected to the drain. This transconductance gm is computed from Eq. 1 by linearizing about the operating point current Idiode as follows:                                                         g              m                        ≡                                          dI                diode                                            dV                gs                                              =                                    I              s                        ⁢            k            ⁢                          xe2x80x83                        ⁢                          exp              ⁡                              (                                  kV                  gs                                )                                                    ⁢                  
                ⁢                              g            m                    =                                    I              s                        ⁢            k            ⁢                          xe2x80x83                        ⁢                          exp              ⁡                              (                                  k                  ⁢                                      1                    k                                    ⁢                  ln                  ⁢                                                            I                      diode                                                              I                      s                                                                      )                                                    ⁢                  xe2x80x83                ⁢                  
                ⁢                  (                      by            ⁢                          xe2x80x83                        ⁢            substituting            ⁢                          xe2x80x83                        ⁢            in            ⁢                          xe2x80x83                        ⁢            Eq            ⁢            .1            ⁢                          xe2x80x83                        ⁢            solved            ⁢                          xe2x80x83                        ⁢            for            ⁢                          xe2x80x83                        ⁢                          V              gs                                )                ⁢                  
                ⁢                              g            m                    =                                                    I                s                            ⁢              k              ⁢                              xe2x80x83                            ⁢                              exp                ⁡                                  (                                      ln                    ⁢                                                                  I                        diode                                                                    I                        s                                                                              )                                                      =                                                            I                  s                                ⁢                k                ⁢                                                      I                    diode                                                        I                    s                                                              =                              kI                diode                                                                        (                  Eq          .                      xe2x80x83                    ⁢          2                )            
Thus, it will be seen that the transconductance gm is proportional to the operating point current Idiode Parasitic capacitances in MOSFET 14 are neglected here, but if included would be incorporated into Cp.
The bandwidth of the photoreceptor circuit is determined by the conductance gm and capacitance Cp, which together form a simple RC low-pass filter of cutoff frequency:                               f          c                =                                            g              m                                      2              ⁢              π              ⁢                              xe2x80x83                            ⁢                              C                p                                              .                                    (                  Eq          .                      xe2x80x83                    ⁢          3                )            
Eq. 2 shows that the conductance gm increases for higher light levels. Thus the cutoff frequency of photoreceptor circuit is higher for higher light levels than for low light levels. For light levels associated with pinhole cameras, the cutoff frequency can be on the order of a fraction of a Hertz. It is noted that according to the model under consideration, the cutoff frequency is independent of photoreceptor area. If the photoreceptor area is doubled, then gm and Cp both double, and thus the cutoff frequency remains constant.
The cutoff frequency is much higher for photoreceptor versions or embodiments using a lens to focus an image of the environment onto the focal plane. This is because a lens gathers a large amount of light and focuses it onto the photoreceptors. Thus the value of gm is increased by up to several orders of magnitude, which increases the bandwidth by an equivalent amount. The main disadvantage of this approach is that a lens is required to gather more light onto the photoreceptors. Of course, in many applications the weight of the lens is not a problem. However, an important optic flow sensor application is in sub-scale xe2x80x9cmicroxe2x80x9d air vehicles of sizes of just several centimeters, and with such vehicles it is not tolerable to have a lens that weighs even a gram.
According to the invention, a photoreceptor array for an optic flow sensor is provided which addresses both of the shortcomings discussed above. The invention provides a way to sample the visual field with rectangular shaped receptive fields without a lens, and without resulting in an extremely low cutoff frequency. Further, the invention is perfectly suited for micro-mechanical systems (MEMS) applications such as those to which reference was made above.
In accordance with the invention, an optic flow sensor is provided which comprises a plurality of photoreceptor circuits, arranged in a linear array, for sensing light received from a visual field and for producing a plurality of photoreceptor signals in accordance with the light sensed; a negative iris disposed between the visual field and the photoreceptor circuits for blocking a portion of the light from the visual field so that the light blocked by the negative iris is not received by the photoreceptor circuits; and optic flow determining means responsive to said photoreceptor signals for determining optic flow in the visual field.
In one preferred embodiment, the negative iris comprises a rectangular light blocking element. Advantageously, the photoreceptor circuits comprise rectangular photoreceptors. Preferably, the rectangular photoreceptors extend parallel to the rectangular iris.
In an alternative preferred embodiment, the negative iris comprises a plurality of light blocking elements. Advantageously, the light blocking elements are rectangular in shape.
According to a further preferred embodiment the negative iris comprises at least one substantially point shaped light blocking element and said photoreceptor circuits comprise rectangular photoreceptors.
In accordance with yet another embodiment, the negative iris comprises a colored iris element for blocking a sub-band of white light.
In a preferred implementation, the sensor further comprises contrast enhancing means for enhancing the contrast levels of the photoreceptor signals. In a first embodiment, the contrast level enhancing means comprises a circuit for receiving the photoreceptor signals, for computing a mean value for the photoreceptor signals and for computing, for each photoreceptor circuit, the difference between the photoreceptor signal for that photoreceptor circuit and said mean. In an alternative embodiment, wherein the optic flow determining means comprises a plurality of feature detectors for receiving the photoreceptor signals, the contrast level enhancing means comprises amplifying means for increasing the amplification of the feature detectors.
In a preferred embodiment, the optic flow sensor is implemented in micro-electromechanical systems technology and the negative iris comprises an etched iris structure.
Other features and advantages of the invention will be set forth in, or will be apparent from, the detailed description of preferred embodiments of the invention which follows.